R Notebook

library(lfe)
library(jtools)
library(stargazer)

Read file generated in Python

country <- "Romania"
year <- 2009

df <- read.csv("data_regressions/Romania_2009_polling_station.csv", encoding="iso8859_16")
df <- df[grepl("Traian", df$candidate), ] # Traian Băsescu
df

model_1 <- lm(rate ~ rate_tudor + rate_antonescu + rate_kelemen + rate_oprescu + rate_basescu, data = df)
model_2 <- lm(rate ~ divisiveness_tudor + divisiveness_antonescu + divisiveness_kelemen + divisiveness_oprescu + divisiveness_basescu, data = df)
model_3 <- lm(rate ~ rate_tudor + rate_antonescu + rate_kelemen + rate_oprescu + rate_basescu + divisiveness_tudor + divisiveness_antonescu + divisiveness_kelemen + divisiveness_oprescu + divisiveness_basescu, data = df)

stargazer(model_1, model_2, model_3, type = "text")

============================================================================================================
                                                        Dependent variable:                                 
                       -------------------------------------------------------------------------------------
                                                               rate                                         
                                   (1)                         (2)                          (3)             
------------------------------------------------------------------------------------------------------------
rate_tudor                       0.536***                                                0.556***           
                                 (0.018)                                                  (0.019)           
                                                                                                            
rate_antonescu                   0.266***                                                0.277***           
                                 (0.006)                                                  (0.006)           
                                                                                                            
rate_kelemen                     0.427***                                                0.398***           
                                 (0.004)                                                  (0.005)           
                                                                                                            
rate_oprescu                     0.432***                                                0.384***           
                                 (0.025)                                                  (0.027)           
                                                                                                            
rate_basescu                     1.047***                                                1.034***           
                                 (0.005)                                                  (0.005)           
                                                                                                            
divisiveness_tudor                                          -2.391***                    -0.566***          
                                                             (0.184)                      (0.089)           
                                                                                                            
divisiveness_antonescu                                      -1.091***                    -0.329***          
                                                             (0.070)                      (0.033)           
                                                                                                            
divisiveness_kelemen                                         0.157***                    0.078***           
                                                             (0.016)                      (0.009)           
                                                                                                            
divisiveness_oprescu                                          -0.251                     1.315***           
                                                             (0.240)                      (0.113)           
                                                                                                            
divisiveness_basescu                                         1.289***                    0.162***           
                                                             (0.058)                      (0.029)           
                                                                                                            
Constant                         0.030***                    0.548***                    0.036***           
                                 (0.003)                     (0.008)                      (0.005)           
                                                                                                            
------------------------------------------------------------------------------------------------------------
Observations                      11,107                      11,107                      11,107            
R2                                0.807                       0.091                        0.812            
Adjusted R2                       0.807                       0.091                        0.812            
Residual Std. Error         0.053 (df = 11101)          0.115 (df = 11101)          0.052 (df = 11096)      
F Statistic            9,305.346*** (df = 5; 11101) 223.174*** (df = 5; 11101) 4,805.735*** (df = 10; 11096)
============================================================================================================
Note:                                                                            *p<0.1; **p<0.05; ***p<0.01

library(minpack.lm)

model_1 <- nlsLM(
    formula = rate ~ const + b_tudor*rate_tudor + b_antonescu*rate_antonescu + b_kelemen*rate_kelemen + b_oprescu*rate_oprescu + b_basescu*rate_basescu,
    data = df, 
    start = list(const=0, b_tudor=0, b_antonescu=0, b_kelemen=0, b_oprescu=0, b_basescu=0),
    lower = c(const=0, b_tudor=0, b_antonescu=0, b_kelemen=0, b_oprescu=0, b_basescu=0),
    upper = c(const=1, b_tudor=1, b_antonescu=1, b_kelemen=1, b_oprescu=1, b_basescu=1),
    #algorithm = "port"
    )

model_2 <- nlsLM(
    formula = rate ~ const + d_tudor*divisiveness_tudor + d_antonescu*divisiveness_antonescu + d_kelemen*divisiveness_kelemen + d_oprescu*divisiveness_oprescu + d_basescu*divisiveness_basescu,
    data = df, 
    start = list(const=0, d_tudor=0, d_antonescu=0, d_kelemen=0, d_oprescu=0, d_basescu=0),
    #lower = c(const=0, b_tudor=0, b_antonescu=0, b_becalli=0, b_kelemen=0, b_oprescu=0, b_basescu=0),
    #upper = c(const=1, b_tudor=1, b_antonescu=1, b_becalli=1, b_kelemen=1, b_oprescu=1, b_basescu=1),
    #algorithm = "port"
    )

model_3 <- nlsLM(
    formula = rate ~ const + b_tudor*rate_tudor + b_antonescu*rate_antonescu + b_kelemen*rate_kelemen + b_oprescu*rate_oprescu + b_basescu*rate_basescu + d_tudor*divisiveness_tudor + d_antonescu*divisiveness_antonescu + d_kelemen*divisiveness_kelemen + d_oprescu*divisiveness_oprescu + d_basescu*divisiveness_basescu,
    data = df, 
    start = list(const=0, b_tudor=0, b_antonescu=0, b_kelemen=0, b_oprescu=0, b_basescu=0, d_tudor=0, d_antonescu=0, d_kelemen=0, d_oprescu=0, d_basescu=0),
    lower = c(const=0, b_tudor=0, b_antonescu=0, b_kelemen=0, b_oprescu=0, b_basescu=0, d_tudor=-Inf, d_antonescu=-Inf, d_kelemen=-Inf, d_oprescu=-Inf, d_basescu=-Inf),
    upper = c(const=1, b_tudor=1, b_antonescu=1, b_kelemen=1, b_oprescu=1, b_basescu=1, d_tudor=Inf, d_antonescu=Inf, d_kelemen=Inf, d_oprescu=Inf, d_basescu=Inf),
    #algorithm = "port"
    )

models <- list(model_1, model_2, model_3)
nlstargazer(models = models)

Summary

Parameters	Model.1	Model.2	Model.3
b_antonescu	0.2468***		0.2629***
b_basescu	1.0000***		1.0000***
b_kelemen	0.4064***		0.3795***
b_oprescu	0.4248***		0.3959***
b_tudor	0.4866***		0.5243***
const	0.0540***	0.5478***	0.0540***
d_antonescu		-1.0911***	-0.3548***
d_basescu		1.2893***	0.1984***
d_kelemen		0.1566***	0.0856***
d_oprescu		-0.2507	1.2528***
d_tudor		-2.3910***	-0.6061***
Residual sum-of-squares	31.33	146.62	30.39

library(minpack.lm)

model_test <- nlsLM(
    formula = rate ~ const + b_tudor*rate_tudor + b_antonescu*rate_antonescu + b_kelemen*rate_kelemen + b_oprescu*rate_oprescu + b_basescu*rate_basescu + d_tudor*divisiveness_tudor*rate_tudor + d_antonescu*divisiveness_antonescu*rate_antonescu + d_kelemen*divisiveness_kelemen*rate_kelemen + d_oprescu*divisiveness_oprescu*rate_oprescu + d_basescu*divisiveness_basescu*rate_basescu,
    data = df, 
    start = list(const=0, b_tudor=0, b_antonescu=0, b_kelemen=0, b_oprescu=0, b_basescu=0, d_tudor=0, d_antonescu=0, d_kelemen=0, d_oprescu=0, d_basescu=0),
    lower = c(const=0, b_tudor=0, b_antonescu=0, b_kelemen=0, b_oprescu=0, b_basescu=0, d_tudor=-Inf, d_antonescu=-Inf, d_kelemen=-Inf, d_oprescu=-Inf, d_basescu=-Inf),
    upper = c(const=1, b_tudor=1, b_antonescu=1, b_kelemen=1, b_oprescu=1, b_basescu=1, d_tudor=Inf, d_antonescu=Inf, d_kelemen=Inf, d_oprescu=Inf, d_basescu=Inf),
    #algorithm = "port"
    )

model_test

Nonlinear regression model
  model: rate ~ const + b_tudor * rate_tudor + b_antonescu * rate_antonescu +     b_kelemen * rate_kelemen + b_oprescu * rate_oprescu + b_basescu *     rate_basescu + d_tudor * divisiveness_tudor * rate_tudor +     d_antonescu * divisiveness_antonescu * rate_antonescu + d_kelemen *     divisiveness_kelemen * rate_kelemen + d_oprescu * divisiveness_oprescu *     rate_oprescu + d_basescu * divisiveness_basescu * rate_basescu
   data: df
      const     b_tudor b_antonescu   b_kelemen   b_oprescu   b_basescu     d_tudor d_antonescu   d_kelemen   d_oprescu   d_basescu 
     0.0312      0.7074      0.3590      0.5810      0.1078      0.9611     -4.6272     -0.8208     -0.6441     16.0734      0.6302 
 residual sum-of-squares: 30.47

Number of iterations to convergence: 3 
Achieved convergence tolerance: 1.49e-08

data1r <- fread("data_output/Romania/2009_first_round.csv.gz", encoding="UTF-8")
data_1r <- data1r[ , .(value = sum(value)), by = candidate]
data_1r$rate <- data_1r$value / sum(data_1r$value)
data_1r[grepl("Traian", data_1r$candidate), "candidate"] <- "Traian Basescu"

candidates <- c(subset(data_1r, rate > 0.02)[["candidate"]])
candidates

[1] "Traian Basescu"       "Crin Antonescu"       "Mircea Geoană"        "Corneliu Vadim Tudor" "Sorin Oprescu"        "Hunor Kelemen"       

coefs <- coef(model_1)
df_model <- data.frame(
  candidate = c("Traian Basescu", "Crin Antonescu", "Corneliu Vadim Tudor", "Sorin Oprescu", "Hunor Kelemen"),
  coef = c(coefs[["b_basescu"]], coefs[["b_antonescu"]], coefs[["b_tudor"]], coefs[["b_oprescu"]], coefs[["b_kelemen"]])
)
df_model <- merge(df_model, data_1r, all=TRUE)
df_model[["1_coef"]] <- 1 - df_model$coef


df_model[["to_candidate_a"]] <- df_model[["rate"]] * df_model[["coef"]]
df_model[["to_candidate_b"]] <- df_model[["rate"]] * df_model[["1_coef"]]

df_model

draw_sankey(df_model, candidates, c("Mircea Geoana", "Traian Basescu"))

[1] "Constantin Ninel Potârcă"
[1] "Mircea Geoana"
[1] FALSE
[1] "Constantin Rotaru"
[1] "Mircea Geoana"
[1] FALSE
[1] "Corneliu Vadim Tudor"
[1] "Mircea Geoana"
[1] FALSE
[1] "Crin Antonescu"
[1] "Mircea Geoana"
[1] FALSE
[1] "George Becali"
[1] "Mircea Geoana"
[1] FALSE
[1] "Gheorghe-Eduard Manole"
[1] "Mircea Geoana"
[1] FALSE
[1] "Hunor Kelemen"
[1] "Mircea Geoana"
[1] FALSE
[1] "Mircea Geoană"
[1] "Mircea Geoana"
[1] FALSE
[1] "Ovidiu Cristian Iane"
[1] "Mircea Geoana"
[1] FALSE
[1] "Remus Cernea"
[1] "Mircea Geoana"
[1] FALSE
[1] "Sorin Oprescu"
[1] "Mircea Geoana"
[1] FALSE
[1] "Traian Basescu"
[1] "Mircea Geoana"
[1] FALSE